In fact, I am reminded of the new word of Brain-Computer Interface (BCI), actually becoming part of the internet of things, where there have been pious talk about bioethics and oversight, all vacuous in the end as big money pushed products out the door already... which deserve another post. The point is that pious talk and political blessings may simply not be enough to save us. Elon Musk is right in broad terms that we are in a scary situation, but the usual types of responses simply will not be enough to save us.
Just a few months back, some leaders in the IT industry asked me to summarize the big picture -- the forest -- as I see it. There are six slides at www.werbos.com/IT_big_picture.
It was really scary for me a few days back to hear Oracle announce a new data platform which, said Bloomberg, "will be ultrasecure because it will use machine learning and deep AI to guarantee security." To me, that sounds like saying "That guy will never catch a virus because he has a great brain." Brains are important, but it is really scary to depend completely on them alone, when you also need a certain kind of immune system and some other such things. Scary as in an urgent matter of life or death. But it was calming to hear that the US actually was able to withstand an all-out cyberattack on our power system from King Kim III; we still have time to develop an integrated secure viable platform before the instabilities and vulnerabilities bring down the whole US to a state lower than Puerto Rico today without assistance.
The new IBM plan, announced by the president of IBM in great, deep interviews easily found on u-tube, seems to be moving us towards the "Stafford Beer" kind of world.
(I owe you a link. Why not try https://www.youtube.com/
At youtube, searching on IBM president IOT AI brings up a huge amount of informaton, really important to these issues, though it may require some background to figure out how it all fits together.) We seem to be moving towards a world where SOMETHING or SOMEONE at the top (a product not only of market but of political forces, more likely Mercer than Sanders) defines a global utility function, and every valve, factory, vehicle and drone on earth is mobilized to intelligently maximize that utility function inexorably, more and more, over all future time. Concern for the less powerful will be even more secondary than it is in recent outcomes in Congress.
An alternative possible pathway would in some ways grow out of what has been developed in the electric power sector in recent years, in the ISO/RTO (Independent System Operator System), and maybe some ebay type experiments. The idea would be to develop an Integrated Market Platform (IMP), which would embody strong unbreakable operating system security and new levels of communication security (and authentication?) , and use a narrow agile variation of distributed ledger technology for additional security of some core currencies (Blockchain 3.0) .
Is there a hope that someone creative enough could save the world by making money developing a new system? I hope so. It's extremely important that someone should try. But no one knows enough yet to just go ahead and start coding.
This morning, as I think over the unsolved design issues, four questions come to the very top of the list:
(1) How does one map CONTROL to OWNERSHIP in a new IT platform? (Again, verbal principles or laws are not enough. "Map" must be like a computer program here.)
A quick naive response would be "That's easy. Ownership for all things, including all devices controlled within the IOT, goes to registered entities which may be humans or may be corporations. Each entity gets to own and run its own instance of the universal IOT platform, which may even have reduced versions available, all subject to strict standards as rigid as any IEEE standards but themselves subject to full open-source automated validation of compliance." But how could many INSTANCES like that all add up, globally, to at least a halfway efficient system? How do they talk to each other? For that, I am reminded of the distributed optimization work pioneered by Marija Ilic (former CMU) and Jean Watson (Sandia) to satisfy the growing need for distributed optimization in the electric power grid, and mathematical extensions worked on by folks like Harley, Venayagamoorthy and myself. A certain kind of automated price system can work, going beyond the usual comparative statics of static market equilibrium and of today's stochastic general equilibrium economics, to something more general.
My wife adds: "Hey, who owns the internet TODAY?" That certainly is something to consider in this design process, the role of standards and how standards affect laws ala net neutrality. The ideas in Tabscott's book on blockchain are relevant, but again, how to make it REAL at the universal platform level?
(2) How does one prevent capture of the system by a ruthless cold minority, creating a kind of value system which grinds down and destroys the very soul of humanity (and its physical existence over time, not forseen by the cabal?). In a world where more and more such cabals are becoming a real problem... I would say we again need open source tools (as we need already for improved security), and market design somehow biased towards full development of all human potential, but realistic about the need to stop exponential population growth and the reality that this need in itself will cause a certain degree of discontent and competition inevitably over time under ANY system. Nice words, no specs? True, and that's why this is a QUESTION, "how to DO this in an IT platform?", and not an answer.
(3) How to make full use of human minds in improving the quality of optimization? For example, is the foresight function performed by RLADP programs (which learn to assess value to the future of commodities or actions in the present) or by human futures trading, or by a mix of the two? WHAT MIX? Mercer has found so far that computers do better than humans in predicting future value, but the underlying mathematics strongly suggests that this is because today's organizations simply don't make full use of what humans can actually do. But unfortunately, that includes a whole lot of lying and gamesmanship and conflict of interest effects which make design of collaborative systems a much deeper problem than IBM's coders seem to understand. (But maybe it's not just them; maybe we need to WORK on that understanding.)
In a brain-like AI system, using RLADP math (see the IEEE book edited by Lewis and Liu) , one can simply set up independent value and prediction networks, using different "currencies"
(derivative signals) to tune themselves, but how could humans feed into it without breaking it?
Actually, as I type this, I have expanded the details so that these three are enough to think about for now. Microeconomics never found an optimal design for an economy of INFORMATION, or for "CONTENT," let alone a reliable, nonscalar system of trust factors to support collaboration. Could the new RLADP math help us with that somehow? Maybe that's a fourth question.
It is truly to have just questions, not answers, when the need for answers and for code which embodies them is so urgent. Every day which passes is bringing us closer to a different, less human-centric world. But posing questions and refusing to forget them has always been the crucial first step in "doing the impossible" (as a few of us have done a few times).
Best of luck,
Paul
==================
This builds on some earlier thinking at
http://drpauljohn.blogspot.com/2017/09/can-natural-markets-solve-our-problems.html
No comments:
Post a Comment